{"id":"90a75b40-e6a2-4124-8750-4f47ec366c02","arxiv_id":"1908.01460","paper_version":3,"verdict":"ACCEPT","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"low","formal_verification":"none","parameter_count":1,"one_line_summary":"A 3GPP-inspired cell-center/cell-edge user pairing enables tractable stochastic geometry analysis of NOMA, showing higher cell sum-rate and effective capacity than OMA.","lead":"This paper models two-user non-orthogonal multiple access (NOMA) in cellular networks by pairing a cell-center user with a cell-edge user, a ranking inspired by 3GPP. It derives rate and delay distributions with stochastic geometry and shows NOMA can improve cell throughput and effective capacity over orthogonal access.","discovery_kind":"new_method","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Central NOMA-vs-OMA claims rest on the Eq. (8) joint-distance approximation, which is not independently validated for the optimized RA comparisons.","rationale":"I identify the same weakest assumption as the reader: Eq. (8). The central claim (NOMA rate-region/CSR advantage, and SEC advantage at high density) would be true if the joint-distance approximation is faithful for CC/CE-conditioned users. It is the least secure point because the typical cell is an area-biased random set; the exact joint distribution of (R_o, R_d) is not derived, and the fitted ρ = 9/7 is calibrated to marginal properties in prior work. The paper deserves credit for explicit approximation statements and for simulation validation in Figs. 2, 4, and 5; those checks show the framework is not merely formal. However, the specific comparisons that constitute the headline result—Fig. 6 rate regions, optimized CSR, and optimized SEC—are produced from the same approximate moments, and Fig. 6's markers are not independent network simulations but brute-force evaluations of the analytic formulation. Thus a residual risk exists that Eq. (8)'s error is amplified under the CC/CE conditioning and at high user density, exactly where the SEC crossover is claimed. Because the paper does not currently provide an end-to-end simulation at those operating points, I would make acceptance conditional on that one check rather than rejecting or accepting unconditionally. This is a targeted verification request, not a claim that the paper is wrong.","tokens_in":26680,"tokens_out":31072,"duration_ms":326052,"concrete_test":"Run an end-to-end Monte Carlo simulation of the actual PPP network at the headline operating points (τ = 0.7, λ = 1, ν ∈ {5, 10}, β = [0, −3] dB, packet sizes T = [50, 50], delay-outage targets O_c = O_e = 0.2), implementing the proposed CC/CE ranking, random scheduling, and the near-optimal RA rules of Section V. Use at least 5 × 10^4 independent cell realizations, and compare the simulated CSR and SEC (with confidence intervals) against the analytic curves in Fig. 6. If the simulated SEC of NOMA at ν = 10 is not above OMA, or the CSR ordering reverses, the Eq. (8) approximation is load-bearing for the central claim and the conclusion needs qualification.","verdict_should_be":"CONDITIONAL","load_bearing_attack":"The load-bearing step is Eq. (8), where the joint density of (R_o, R_d) for a uniform point in the typical cell is replaced by the two-nearest-point joint density of a PPP of intensity ρλ, ρ = 9/7. This single approximation feeds Lemma 1 and Theorem 1, and therefore every mean rate, rate CDF, delay bound, CSR, and SEC expression in Sections IV–VI. The exact typical-cell law is area-biased; the true dependence between R_o and R_d, and its behavior under the CC/CE conditions R_o ≤ τR_d and R_o > τR_d, is more complex than a PPP two-nearest-point law. The constant ρ = 9/7 is imported from prior marginal-distance work, and no evidence is given that it also matches the joint law or the conditional tails that define CC and CE users. The agreement in Figs. 2, 4, and 5 is reassuring but is shown for selected parameter points; Fig. 6, which contains the headline rate-region/CSR/SEC comparisons, has no direct network-simulation markers for the optimized RA policies—its 'optimal' curves are brute-force evaluations of the same approximate expressions. Consequently, if Eq. (8) is biased in a way that is mild for the plotted means but larger for the CC/CE conditional tails, the qualitative conclusion that NOMA outperforms OMA at higher user density could change.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"This paper analyzes downlink two-user NOMA in a Poisson cellular network, pairing one cell-center (CC) user and one cell-edge (CE) user, where the CC/CE regions are defined by whether the service link distance Ro is below or above tau times the distance Rd to the dominant interfering BS. The main technical contributions are approximate moments of the meta distributions for CC/CE users under NOMA and OMA (Theorem 1 and Corollary 1), beta-approximated meta distributions, load-aware transmission-rate CDFs and mean-delay CCDFs (Theorems 2 and 3 along with corollaries), and resource-allocation formulations that maximize cell sum-rate for non-real-time traffic and sum effective capacity for real-time traffic (Section V). The numerical section validates the analytical expressions against Monte Carlo simulations and reports that NOMA improves the rate region, cell sum-rate, and (at higher user density) sum effective capacity relative to OMA.","tokens_in":26925,"tokens_out":9478,"duration_ms":100436,"significance":"If the quantitative conclusions hold, the paper makes a useful advance: it moves NOMA analysis from typical-user and distance-based ranking to a typical-cell formulation that respects the correlation between paired users' distances and the dependence on the dominant interferer, and it connects the meta-distribution analysis to load-aware rate and delay metrics. The same approximation chain is applied symmetrically to NOMA and OMA, which reduces the risk that the NOMA-vs-OMA comparison is forced by construction. Strengths include the Monte Carlo verification of the meta-distribution moments (Fig. 2), the rate and delay CDFs (Figs. 4 and 5), and the comparison of the proposed near-optimal RA with brute-force solutions (Fig. 6).","major_comments":[{"comment":"The joint density approximation in Eq. (8), with the correction factor rho=9/7 taken from the authors' prior work [38], is the foundation of Lemma 1, Theorem 1, and hence of every rate, delay, CSR, and SEC result in Sections IV-VI. The paper validates the downstream metrics only at selected parameter points in Figs. 2, 4, and 5; Fig. 6, which contains the headline rate-region, CSR, and SEC comparisons, has no direct network-simulation markers for the optimized RA policies, and the 'exact optimal' curves are brute-force evaluations of the same approximate expressions. Because Lemma 1 conditions on the events Ro <= tau*Rd and Ro > tau*Rd, an error in the conditional tails of Eq. (8) could affect the NOMA and OMA comparisons unequally. Please add a direct simulation check of the joint and conditional distance statistics, and Monte Carlo markers for at least the rate-region and the optimized CSR/SEC curves over the range of nu and tau used in Fig. 6.","section":"Section III, Eq. (8); Section VI, Fig. 6"},{"comment":"The independence of the CC/CE load and the conditional success probability is used to factor expectations in Theorems 2 and 3. The paper states that the numerical results demonstrate the accuracy of this assumption, but the supporting figures use one default parameter set (plus tau=0.8) and do not quantify the error over the nu, beta, and tau ranges that drive the SEC comparison in Fig. 6 (right). Since the SEC conclusion at higher user density depends on the load distributions and Assumption 1, please report the maximum deviation of the analytical rate/delay CDFs from simulation across the parameter ranges of Fig. 6, or otherwise bound the error introduced by the independence assumption.","section":"Section II-D, Assumption 1; Section IV, Theorems 2-3"}],"minor_comments":[{"comment":"In the CE-OMA moment, the denominator is written with chi_e; since OMA has no power-splitting parameter, it should read beta_e, matching the argument of Z_b and the CC-OMA expression.","section":"Corollary 1, Eq. (19)"},{"comment":"The phrase 'zero-truncated Poisson distributions with means nu|Vxc|' is imprecise: the mean of the zero-truncated Poisson with parameter nu*a is nu*a/(1-exp(-nu*a)). The subsequent formulas (24)-(25) correctly use the zero-truncated pmf, so please adjust the wording.","section":"Section II-A, after Eq. (2)"},{"comment":"The statement that the CC and CE probabilities are tau^2 and 1-tau^2 should be labeled as consequences of the approximate joint density (8) rather than as exact PV-cell probabilities.","section":"Lemma 1"},{"comment":"'CSEOMA' appears to be a typo for 'CSROMA'.","section":"Section V-A.2"}],"recommendation":"major_revision","confidential_remarks":"The paper's quantitative claims rest on the rho=9/7 joint-distance approximation imported from the authors' own prior work. The revision should include a robustness check with respect to rho (e.g., rho=1 and rho=1.3) in the optimized RA comparisons, in addition to the simulation markers requested in the major comments. This is a correctness-risk concern rather than an allegation of improper fitting; applying the same approximation chain to both OMA and NOMA does mitigate circularity."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Here's my take.\n\nWhat you should know: this paper does something new and useful. Instead of the usual typical-user analysis, it analyzes a user cluster in the typical cell, pairing a cell-center (CC) and cell-edge (CE) user defined by the ratio of distances to the serving and dominant interfering BSs. That ranking depends on inter-cell interference, not just serving-link distance, which is a real improvement over most stochastic-geometry NOMA work. The moments of the meta distributions for both users (Theorem 1) are closed-form, and the paper is careful about validating them against Monte Carlo.\n\nThere's a lot to credit. The authors state their assumptions explicitly, the load modeling with zero-truncated Poisson and gamma area fits is standard but done cleanly, and the NOMA-vs-OMA comparison is framed fairly: both systems are analyzed with the same machinery, so the conclusion isn't forced by construction. The numerics in Figs. 2, 4, and 5 show good agreement across a reasonable parameter range.\n\nNow the soft spots, in proportion.\n\nThe whole analysis rests on Eq. (8): the joint pdf of the service-link distance and the dominant-interference distance is replaced by the two-nearest-point joint density in a PPP with density ρλ, with ρ=9/7 imported from the authors' earlier work. That approximation feeds directly into Lemma 1 and Theorem 1, and hence into every rate, delay, CSR, and SEC expression. The paper validates it for the plotted means and distributions, but the CC/CE conditions (Ro ≤ τRd and its complement) are exactly the conditional tails where a two-point approximation could miss. The stress-test note is right that Fig. 6—the headline rate-region/CSR/SEC comparison—has no network-simulation markers for the optimized RA policies; the 'optimal' curves are brute-force evaluations of the same approximate expressions. That is a genuine gap in the evidence. It doesn't kill the paper, because both NOMA and OMA inherit the same approximation, and the qualitative result (NOMA beats OMA at higher density) is plausible. But a referee should ask for simulation verification of at least the optimized RA points.\n\nOne more small thing: Assumption 1 (load-success independence) is another layer, but it's a standard type of assumption in this literature and the authors test it indirectly.\n\nBottom line: this is a solid paper for the stochastic-geometry NOMA audience. It advances the modeling toolkit and is honest about its approximations. It deserves peer review, with the joint-distance approximation and the Fig. 6 simulations as the focus.\n\nMy recommendation: send it to review. I'd want a referee to rerun the Fig. 6 comparisons with actual network simulations before publication.","headline":"A solid typical-cell stochastic-geometry analysis of two-user NOMA with a 3GPP-inspired ranking; the load-bearing joint-distance approximation needs an extra simulation check on the optimized RA comparisons.","tokens_in":27466,"tokens_out":2583,"would_cite":true,"duration_ms":26082,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["60D05","60G55","94A05"],"pacs":[],"model":"deepseek-v4-flash","headline":"This paper claims that pairing one cell-center user with one cell-edge user, dividing them by the ratio of serving-to-dominant-interferer distances, makes downlink NOMA analytically tractable and lets it beat OMA on rate region, cell…","keywords":["non-orthogonal multiple access","user ranking","meta distribution","stochastic geometry","Poisson point process","cell-center cell-edge regions","effective capacity","resource allocation"],"falsifier":"Simulate a Poisson-Voronoi cellular network with the paper's cell-center/cell-edge partition and directly measure the joint distribution of the two distances. If the empirical conditional distance distributions deviate from Lemma 1's formulas beyond Monte Carlo error, or if the empirical first and second moments of the conditional success probability disagree with Eq. (17), the central approximation is falsified.","tokens_in":26429,"feed_emoji":"📶","tokens_out":11046,"duration_ms":101963,"temperature":0.7,"pith_summary":"This paper tries to show that a particular pairing rule makes non-orthogonal multiple access (NOMA) worth using in dense cellular networks. Instead of pairing the users closest to a base station, it pairs one user from the cell-center region with one from the cell-edge region, where the dividing line is set by comparing the path loss to the serving base station with the path loss to the strongest interfering base station. For this pairing, the paper derives closed-form moment formulas for the meta distribution, meaning the network-wide distribution of a user's success probability given the base-station locations, and from them the transmission-rate and packet-delay distributions under random scheduling. The numerical payoff is that NOMA yields a larger rate region and higher cell sum-rate than orthogonal access, and better sum effective capacity at higher user density. That matters because earlier distance-based rankings placed paired users in unrealistic locations and did not isolate the channel-quality gap NOMA needs.","feed_headline":"Interference-aware pairing, not distance, makes NOMA beat OMA","feed_subtitle":"Pairing cell-center with cell-edge users raises rate region, cell sum-rate, and delay-aware capacity.","key_machinery":"The load-bearing machinery is the approximate joint probability density of the two link distances that define the pairing: the service distance $R_o$ to the serving base station and the distance $R_d$ to the dominant interferer. Eq. (8) approximates this joint density by $(2\\pi\\rho\\lambda)^2 r_o r_d \\exp(-\\pi\\rho\\lambda r_d^2)$ for $r_d \\ge r_o \\ge 0$, with correction factor $\\rho = 9/7$; conditioning on $R_o \\le \\tau R_d$ for center users and $R_o > \\tau R_d$ for edge users turns this into the conditional distance laws in Lemma 1. These laws feed Theorem 1's moment formulas for the meta distribution, and the same calibrated joint density is what keeps the later rate, delay, and resource-allocation analysis closed form.","core_discovery":"The central claim is that the cell-center/cell-edge ranking captures the correlation between the two paired users' link qualities that earlier distance-based rankings threw away, and that this correlation is exactly what the analysis can exploit. A user in the typical cell is classified as a center user if its service distance $R_o$ satisfies $R_o \\le \\tau R_d$, where $R_d$ is the distance to the dominant interfering base station, and as an edge user otherwise. With the joint density of these two distances approximated as in Eq. (8), Theorem 1 gives the $b$-th moments of the meta distributions as $M_c^b(\\chi_c) = \\frac{\\rho^2}{\\tau^2}\\int_0^{\\tau^2} \\frac{(\\rho + v Z_b(\\chi_c,v))^{-2}}{(1+\\chi_c v^{1/\\delta})^b}\\,dv$ and $M_e^b(\\chi_e) = \\frac{\\rho^2}{1-\\tau^2}\\int_{\\tau^2}^{1} \\frac{(\\rho + v Z_b(\\chi_e,v))^{-2}}{(1+\\chi_e v^{1/\\delta})^b}\\,dv$, where $\\chi_c = \\max\\{\\beta_c/\\theta,\\ \\beta_e/(1-\\theta(1+\\beta_e))\\}$, $\\chi_e = \\beta_e/(1-\\theta(1+\\beta_e))$, and $\\delta = 2/\\alpha$. These moments feed $\\beta$-distribution approximations of the meta distributions, load-aware rate and mean-delay distributions, and near-optimal resource allocations for maximizing cell sum-rate and sum effective capacity. On the paper's own numerical evidence, NOMA with this ranking gives a strictly larger rate region than OMA, a higher maximum cell sum-rate, and a higher sum effective capacity at high user density.","pith_inferences":["If Eq. (8) is as accurate as the paper's simulations suggest, the same center/edge construction with $N-1$ thresholds should yield a moment theory for $N$-user NOMA; the two-user formulas in Theorem 1 are the seed of that generalization.","The partition rule is defined solely by a serving-to-dominant-interferer distance ratio, so it can be transplanted to other schemes that divide users by quality of service, such as soft frequency reuse, giving those schemes the same stochastic-geometry machinery.","The delay results are upper bounds built on saturated interfering queues; a simulation that lets interfering queues be non-saturated would show how much conservatism remains in the real-time-traffic conclusions.","The optimal-power rule $\\theta^* = \\min(\\theta_e, \\hat{\\theta})$ could be implemented online if the moment functions are precomputed, turning an analytic result into a low-complexity scheduler."],"forward_implications":["For scheduling fractions up to about $\\eta = 0.72$ in the paper's setting, both the center and edge users get higher mean transmission rates under NOMA than under OMA, so the cell-level gain is not obtained by starving the edge user.","The near-optimal NOMA power split is simple: give the edge layer just enough power to meet its rate target and the rest to the center layer up to the balancing point $\\hat{\\theta}$.","With non-real-time traffic, maximizing cell sum-rate under minimum-rate constraints favors NOMA over OMA at every user density tested, and the feasible region shrinks as density grows for both.","With real-time traffic, NOMA's sum effective capacity exceeds OMA's at higher user density when the SIR thresholds are low, because concurrent transmission raises the packet service rate.","The beta approximations make the rate and delay distributions accurate enough for system-level evaluation without numerically inverting the meta distribution, which is what makes the allocation problems tractable."],"supporting_citations":[{"why":"Supplies the principle that NOMA gains require pairing users with distinct link qualities, which motivates the cell-center/cell-edge ranking.","marker":"[4]"},{"why":"Defines the meta distribution and the moment-based analysis that Theorem 1 extends to the typical cell.","marker":"[19]"},{"why":"Inspires the cell-center/cell-edge classification by serving versus dominant-interferer power ratio.","marker":"[22]"},{"why":"Defines effective capacity, the quality-of-service metric used for the real-time traffic resource allocation.","marker":"[32]"},{"why":"Provides the typical-cell framework and the correction-factor joint-distance approximation on which Eq. (8) rests.","marker":"[38]"},{"why":"Introduces the load and success-probability independence assumption and the load-aware analysis used in Theorems 2 and 3.","marker":"[40]"},{"why":"Supplies the Type I user point process construction used to model cell-center and cell-edge user locations.","marker":"[41]"},{"why":"Gives the gamma approximation for Poisson-Voronoi cell area used to derive the cell-center and cell-edge load distributions.","marker":"[51]"}],"fun_headline_variants":["Interference-aware user ranking makes NOMA beat OMA","NOMA gains stem from link-quality correlation in user clusters","3GPP-inspired ranking unlocks NOMA downlink gains","Cell-center/edge pairing boosts NOMA rate region and capacity","NOMA beats OMA with interference-based user pairing"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The entire calculation leans on approximating the joint statistics of the distances to the serving and dominant interfering base stations by those of the two nearest points in a random pattern of base stations whose density is scaled by $9/7$; if real center or edge users do not follow that approximation, the rates, delays, and allocations would shift.","fun_headline_variants_meta":{"raw":{"variants":["Interference-aware user ranking makes NOMA beat OMA","NOMA gains stem from link-quality correlation in user clusters","3GPP-inspired ranking unlocks NOMA downlink gains","Cell-center/edge pairing boosts NOMA rate region and capacity","NOMA beats OMA with interference-based user pairing"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000843,"raw_usage":{"total_tokens":3800,"prompt_tokens":1200,"completion_tokens":2600,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":816,"completion_tokens_details":{"reasoning_tokens":2519}},"tokens_in":816,"tokens_out":2600,"duration_ms":16971,"temperature":1.0,"reasoning_tokens":2519,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-14T15:12:50.188559+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Simulate a Poisson-Voronoi cellular network with the paper's cell-center/cell-edge partition and directly measure the joint distribution of the two distances. If the empirical conditional distance distributions deviate from Lemma 1's formulas beyond Monte Carlo error, or if the empirical first and second moments of the conditional success probability disagree with Eq. (17), the central approximation is falsified.","supporting_citations":[{"cited_title":"Impact of user pairing on 5G non- orthogonal multiple-access downlink transmissions,","cited_arxiv_id":null,"evidence_quote":"Supplies the principle that NOMA gains require pairing users with distinct link qualities, which motivates the cell-center/cell-edge ranking."},{"cited_title":"The meta distribution of the SIR in Poisson bipolar and cellular networks,","cited_arxiv_id":null,"evidence_quote":"Defines the meta distribution and the moment-based analysis that Theorem 1 extends to the typical cell."},{"cited_title":"Self- organizing interference management for LTE,","cited_arxiv_id":null,"evidence_quote":"Inspires the cell-center/cell-edge classification by serving versus dominant-interferer power ratio."},{"cited_title":"Effective capacity: a wireless link model for support of quality of service,","cited_arxiv_id":null,"evidence_quote":"Defines effective capacity, the quality-of-service metric used for the real-time traffic resource allocation."},{"cited_title":"Downlink analysis for the typical cell in Poisson cellular networks,","cited_arxiv_id":null,"evidence_quote":"Provides the typical-cell framework and the correction-factor joint-distance approximation on which Eq. (8) rests."},{"cited_title":"Stochastic geometry-based uplink analysis of massive MIMO systems with fractional pilot reuse,","cited_arxiv_id":null,"evidence_quote":"Introduces the load and success-probability independence assumption and the load-aware analysis used in Theorems 2 and 3."},{"cited_title":"User point processes in cellular networks,","cited_arxiv_id":null,"evidence_quote":"Supplies the Type I user point process construction used to model cell-center and cell-edge user locations."},{"cited_title":"Statistical distributions of Poisson V oronoi cells in two and three dimensions,","cited_arxiv_id":null,"evidence_quote":"Gives the gamma approximation for Poisson-Voronoi cell area used to derive the cell-center and cell-edge load distributions."}],"review_version":1}